Learning to predict relapse in invasive ductal carcinomas based on the subcellular localization of junctional proteins

Department of Computing Science, University of Alberta, 359 Athabasca Hall, Edmonton, AB T6G 2E8, Canada.
Breast Cancer Research and Treatment (Impact Factor: 3.94). 09/2009; 121(2):527-38. DOI: 10.1007/s10549-009-0557-0
Source: PubMed


The complexity of breast cancer biology makes it challenging to analyze large datasets of clinicopathologic and molecular attributes, toward identifying the key prognostic features and producing systems capable of predicting which patients are likely to relapse. We applied machine-learning techniques to analyze a set of well-characterized primary breast cancers, which specified the abundance and localization of various junctional proteins. We hypothesized that disruption of junctional complexes would lead to the cytoplasmic/nuclear redistribution of the protein components and their potential interactions with growth-regulating molecules, which would promote relapse, and that machine-learning techniques could use the subcellular locations of these proteins, together with standard clinicopathological data, to produce an efficient prognostic classifier. We used immunohistochemistry to assess the expression and subcellular distribution of six junctional proteins, in addition to a panel of eight standard clinical features and concentrations of four "growth-regulating" proteins, to produce a database involving 36 features, over 66 primary invasive ductal breast carcinomas. A machine-learning system was applied to this clinicopathologic dataset to produce a decision-tree classifier that could predict whether a novel breast cancer patient would relapse. We show that this decision-tree classifier, which incorporates a combination of only four features (nuclear alpha- and beta-catenin levels, the total level of PTEN and the number of involved axillary lymph nodes), is able to correctly classify patient outcomes essentially 80% of the time. Further, this classifier is significantly better than classifiers based on any subgroup of these 36 features. This study demonstrates that autonomous machine-learning techniques are able to generate simple and efficient decision-tree prognostic classifiers from a wide variety of clinical, pathologic and biomarker data, and unlike other analytic methods, suggest testable biologic relationships among explicitly identified key variables. The decision-tree classifier resulting from these analytic methods is sufficiently simple and should be widely applicable to a spectrum of clinical cancer settings. Further, the subcellular distribution of junctional proteins, which influences growth regulatory pathways involved in locoregional and metastatic relapse of breast cancer, helped to identify which patients would relapse while their total concentration did not. This emphasizes the need to evaluate the subcellular distribution of junctional proteins in assessing their contribution to tumor progression.

16 Reads
  • Source
    • "For instance, the expression of a β-catenin mutant with an abnormally high stability has been shown to induce breast adenocarcinomas in a transgenic mouse model [11]. By immunohistochemistry, the expression of βcatenin in breast cancer (reported to be up to 60% of the cases) has been reported to significantly correlate with a poor prognosis or relapse in breast cancer patients in previous studies [12] [13] [14]. A few previous studies have shed light to the mechanisms underlying the relatively "
    [Show abstract] [Hide abstract]
    ABSTRACT: The expression of β-catenin detectable by immunohistochemistry has been reported to be prognostically important in breast cancer. In this study, we investigated the mechanism by which β-catenin is regulated in breast cancer cells. Our analysis of the gene promoter of β-catenin revealed multiple putative STAT3 binding sites. In support of the concept that STAT3 is a transcriptional regulator for β-catenin, results from our chromatin immunoprecipitation studies showed that STAT3 binds to two of the three potential STAT3-binding sites in the gene promoter of β-catenin (-856 and -938). Using our generated MCF-7 cell clones that carry an inducible STAT3C construct, we found that the expression levels of STAT3C significantly correlated with the transcriptional activity of β-catenin. Similar observations were made when we subjected two breast cancer cell lines (MCF-7 and BT-474) to STAT3 knock-down or transient gene transfection of STAT3C. Using immunohistochemistry, we found that pSTAT3 and β-catenin significantly correlated with each other (p=0.003, Fisher's exact test) in a cohort of primary breast tumors (n=129). To conclude, our results support the concept that STAT3 upregulates the protein expression and transcriptional activity of β-catenin in breast cancer, and these two proteins may cooperate with each other in exerting their oncogenic effects in these tumors.
    International journal of clinical and experimental pathology 01/2010; 3(7):654-64. · 1.89 Impact Factor
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: In the last 15 years, there has been a flourishing of research into the neural basis of reinforcement learning, drawing together insights and findings from psychology, computer science, and neuroscience. This remarkable confluence of three fields has yielded a growing framework that begins to explain how animals and humans learn to make decisions in real time. Mastering the literature in this sub-field can be quite daunting as this task can require mastery of at least three different disciplines, each with its own jargon, perspectives, and shared background knowledge. In this chapter, the authors attempt to make this fascinating line of research more accessible to researchers in any of the constitutive sub-disciplines. To this end, the authors develop a primer for reinforcement learning in the brain that lays out in plain language many of the key ideas and concepts that underpin research in this area. This primer is em- bedded in a literature review that aims not to be comprehensive, but rather representative of the types of questions and answers that have arisen in the quest to understand reinforcement learning and its neural substrates. Drawing on the basic findings in this research enterprise, the authors conclude with some speculations about how these developments in computational neuroscience may influence future developments in Artificial Intelligence.
    Computational Neuroscience for Advancing Artificial Intelligence: Models, Methods and Applications, Edited by Eduardo Alonso, Esther Mondragon, 01/2011: chapter 6: pages 111-144; IGI Global.
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: Nucleophosmin (NPM) is a nucleolar phosphoprotein that is involved in many cellular processes and has both oncogenic and growth suppressing activities. NPM is localized primarily in nucleoli but shuttles between the nucleus and the cytoplasm, and sustained cytoplasmic distribution contributes to its tumor promoting activities. Plakoglobin (PG, γ-catenin) is a homolog of β-catenin with dual adhesive and signaling functions. These proteins interact with cadherins and mediate adhesion, while their signaling activities are regulated by association with various intracellular partners. Despite these similarities, β-catenin has a well-defined oncogenic activity, whereas PG acts as a tumor/metastasis suppressor through unknown mechanisms. Comparison of the proteomic profiles of carcinoma cell lines with low- or no PG expression with their PG-expressing transfectants has identified NPM as being upregulated upon PG expression. Here, we examined NPM subcellular distribution and in vitro tumorigenesis/metastasis in the highly invasive and very low PG expressing MDA-MB-231 (MDA-231) breast cancer cells and their transfectants expressing increased PG (MDA-231-PG) or NPM shRNA (MDA-231-NPM-KD) or both (MDA-231-NPM-KD+PG). Increased PG expression increased the levels of nucleolar NPM and coimmunoprecipitation studies showed that NPM interacts with PG. PG expression or NPM knockdown decreased the growth rate of MDA-231 cells substantially and this reduction was decreased further in MDA-231-NPM-KD+PG cells. In in vitro tumorigenesis/metastasis assays, MDA-231-PG cells showed substantially lower and MDA-231-NPM-KD cells substantially higher invasiveness relative to the MDA-231 parental cells, and the co-expression of PG and NPM shRNA led to even further reduction of the invasiveness of MDA-231-PG cells. Furthermore, examination of the levels and localization of PG and NPM in primary biopsies of metastatic infiltrating ductal carcinomas revealed coordinated expression of PG and NPM. Together, the data suggest that PG may regulate NPM subcellular distribution, which may potentially change the function of the NPM protein from oncogenic to tumor suppression.
    Oncogenesis 03/2012; 1(3):e4. DOI:10.1038/oncsis.2012.4 · 3.95 Impact Factor